Opening the research
Opening the research
The rules, the differences, and the outcomes—including the ideas that did not work.
The main strategy follows daily trends, with positions typically lasting days to weeks. It is long-only, uses equal position sizes, and holds at most one position per coin.
The January 2025–September 2026 unseen-data test produced 243 trades, a 33.7% win rate, +1.5% average net return per trade, and a 1.32 profit factor. Average winners were about +19%; average losers about −7.3%.
The 95% confidence interval for average return was −2.9% to +6.3%. It includes zero. There were only 33 independent entry-weeks, and development results leaned heavily on the 2020–2021 bull market. The result is inconclusive, not proof of a durable edge.
v006 applies the same daily rules to 36 coins, including newer listings such as HYPE, SUI, ENA, and TAO. It runs alongside v005 rather than replacing it.
Many newer coins lack enough history for a meaningful backtest. This version is evaluated on live paper results from 9 October 2026. A coin requires 20 daily closes before it can become tradable.
v008 holds two halves on 37 Binance futures. The trend half averages the direction of each coin’s 20-, 60- and 120-day returns and goes long or short, sized by volatility and rebalanced daily. The momentum half ranks coins by their last-month return each week, buys the top third and shorts the bottom third in equal amounts, so it is roughly market-neutral. Each half aims for 20% annual volatility, with at most 25% in one coin and 2× gross exposure.
These rules come from earlier research that has already seen every past period, so no backtest can test them fairly. v008 is judged on live paper results only, from 18 October 2026. After 90 days it passes only if its Sharpe ratio is above 0.5, both halves are positive, the drawdown stays under 25%, and it beats holding the same coins on a risk-adjusted basis. Otherwise it is marked failed, not retuned.
A replay on 2021–2026 data, which is not a fair test, made about 10% a year with a 21% worst drawdown, mostly from the trend half. It was down about 7% in 2026 and charges no funding costs. v008 runs on paper and sends no signals.
Only v005 and v006 send signals. As of 10 October 2026 they have sent one: an ATOM entry at 2.066 after that day’s close, marked about 6.7% below entry later the same day. It stays open until the exit rule closes it.
One open trade says nothing about the strategy. In testing about two trades in three lost and the average winner was roughly +19%, so a losing start is the expected case, not a surprise. Every live signal is added to the record, win or lose.
Failed strategies remain part of the record. They are not quietly retuned and presented as fresh successes.
v003, v004 and v007 still run on paper so their forward behavior can be compared with the tests. They send no signals.